Essay1949
Translation
Warren Weaver
Proposes treating translation as a cryptographic decoding problem, arguing that a text in one language can be seen as the same message encoded, and that wartime code-breaking statistics could help recover it.
Computational Linguistics
Language processed by machine.
8 topics · 13 curated works
Topics
- 01Foundations & Overviews
- 02Corpus Linguistics2
- 03Parsing (Linguistic)1
- 04Machine Translation (Linguistic)2
- 05Speech Recognition2
- 06Language Models as Linguistic Evidence2
- 07Annotation & Treebanks2
- 08Low-Resource Languages2
Reading in Computational Linguistics
13A way in
Start here
No prior grounding assumed.
Translation
Warren Weaver · 1949
Proposes treating translation as a cryptographic decoding problem, arguing that a text in one language can be seen as the same message encoded, and…
+1 more at this level
Then
Assumes you know the vocabulary.
A Standard Sample of Present-Day Edited American English, for Use with Digital Computers
W. Nelson Francis · 1964
Describes the compilation of the Brown Corpus, a million-word sample of edited prose built to a fixed sampling design, establishing the electronic…
+3 more at this level
Go deeper
Primary sources and full treatments.
A Statistical Approach to Machine Translation
Brown et al. · 1990
Argues that translation equivalences can be learned automatically from a large aligned bilingual corpus alone, without any hand-written bilingual…
+6 more at this level
12 of 13 works
Essay2005
Automatic Speech Recognition – A Brief History of the Technology Development
B. H. Juang & Lawrence R. Rabiner
Traces automatic speech recognition from rigid template-matching methods to statistical hidden Markov modelling, arguing the shift to a probabilistic framework in the 1980s, not any single algorithm, is what made large-vocabulary continuous recognition practical.
link checked 17 Sept 2026Report1964
A Standard Sample of Present-Day Edited American English, for Use with Digital Computers
W. Nelson Francis
Describes the compilation of the Brown Corpus, a million-word sample of edited prose built to a fixed sampling design, establishing the electronic corpus as a reusable, shared object for studying a language empirically.
FreeIntermediate
Paper1990
Word Association Norms, Mutual Information, and Lexicography
Kenneth Ward Church & Patrick Hanks
Argues word association strength can be measured objectively from corpus co-occurrence statistics using mutual information, replacing a lexicographer's intuitive judgement of collocation with a repeatable count over text.
link checked 17 Sept 2026FreeIntermediate
Paper2016
Universal Dependencies v1: A Multilingual Treebank Collection
Nivre et al.
Presents a single dependency-annotation scheme applied consistently across many languages, arguing that a shared set of grammatical relations makes syntactic annotation comparable across languages instead of being redesigned for each one.
FreeIntermediate
Paper2020
The State and Fate of Linguistic Diversity and Inclusion in the NLP World
Joshi et al.
Categorises the world's languages by how much labelled data and research attention NLP has given them and shows the field's progress has concentrated on a handful of languages, leaving most of linguistic diversity untouched.
FreeIntermediate
Paper1990
A Statistical Approach to Machine Translation
Brown et al.
Argues that translation equivalences can be learned automatically from a large aligned bilingual corpus alone, without any hand-written bilingual dictionary or grammar rules.
Paper1990
Improved Acoustic Modeling for Continuous Speech Recognition
Lee, Giachin, Rabiner, Pieraccini & Rosenberg
Extends the hidden-Markov-model approach to continuous speech recognition Jelinek's group pioneered with context-dependent phone models and better parameter smoothing, and measures the resulting drop in word-error rate.
Paper1993
Building a Large Annotated Corpus of English: The Penn Treebank
Mitchell P. Marcus, Beatrice Santorini & Mary Ann Marcinkiewicz
Describes the construction of the Penn Treebank, a corpus of hand-annotated syntactic structure, which became the shared benchmark that let parsers built on entirely different theories be compared against each other.
link checked 17 Sept 2026Paper2003
Head-Driven Statistical Models for Natural Language Parsing
Michael Collins
Shows that a parser conditioned on lexical head-to-head dependencies, trained on a treebank, outperforms parsers built from hand-written grammar rules.
link checked 17 Sept 2026Paper2009
Linguistically Naïve != Language Independent: Why NLP Needs Linguistic Typology
Emily M. Bender
Argues that language processing systems built and tested only on English quietly encode English's grammatical assumptions, and that engineers need at least basic linguistic typology to notice when a method will not transfer.
link checked 17 Sept 2026Paper2016
Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies
Tal Linzen, Emmanuel Dupoux & Yoav Goldberg
Tests whether a recurrent network trained only on raw text learns long-distance subject-verb agreement, and uses the cases where it fails as evidence about what statistical learning alone can and cannot recover of syntax.
link checked 17 Sept 2026In order written
1949 – 2020- 1949TranslationWarren Weaver
- 1990Word Association Norms, Mutual Information, and LexicographyKenneth Ward Church & Patrick Hanks
- 1990A Statistical Approach to Machine TranslationBrown et al.
- 1990Improved Acoustic Modeling for Continuous Speech RecognitionLee, Giachin, Rabiner, Pieraccini & Rosenberg
- 1993Building a Large Annotated Corpus of English: The Penn TreebankMitchell P. Marcus, Beatrice Santorini & Mary Ann Marcinkiewicz
- 2003Head-Driven Statistical Models for Natural Language ParsingMichael Collins
- 2005Automatic Speech Recognition – A Brief History of the Technology DevelopmentB. H. Juang & Lawrence R. Rabiner
- 2016Universal Dependencies v1: A Multilingual Treebank CollectionNivre et al.
- 2016Assessing the Ability of LSTMs to Learn Syntax-Sensitive DependenciesTal Linzen, Emmanuel Dupoux & Yoav Goldberg
- 2019What Does BERT Look at? An Analysis of BERT's AttentionClark, Khandelwal, Levy & Manning
Also covered elsewhere
This subject genuinely sits in more than one domain. These fields approach the same ground with different methods.